The dynamics of sovereign credit default swap and bond markets: empirical evidence from the 2001 to 2007 period

Abstract

This article evaluates the dynamic relationship between sovereign Credit Default Swap (CDS) and bond markets over the period 2001 to 2007 across 30 emerging markets. Our results suggest that the bond markets play a significant role in the price discovery process, which is in contrast with the corporate studies. We also show that the CDS markets play a more dominant role in lead–lag relationships compared to earlier studies on the sovereign credit markets.

Keywords:

• sovereign
• credit default swap
• bond spread
• price discovery

JEL Classification:

• F36
• G14

Acknowledgements

We thank JP Morgan for the EMBIG spread (monthly) data for 32 countries.We are grateful to Nandu Nayar, Vladimir Dobric and Parveen Gupta for providing their input. We also received useful input from two anonymous referees and the participants at presentations held at the annual meetings of the Eastern Economic Association (New York City, 2009), the Eastern Finance Association (Washington, DC, 2009), the European Financial Management Association (Milan, 2009) and the Financial Management Association (Reno, NV, 2009). Special thanks to JP Morgan and Markit Group for providing the data.

All errors are the responsibility of the authors.

Notes

¹Except Russia and Croatia where the CDS quotes are Euro denominated. All quotes are characterized by a Complete Restructuring (CR) clause (old restructuring) and a senior tier, and they are acquired from composite contributors.

²The exact methodology on how they calculated the EMBIG spreads is proprietary information and kept confidential by JP Morgan. However, if the calculation is using a monthly average versus last day of the month calculation, our analysis might suffer from ‘asynchronous data bias’. In order to control for this bias and robustness purposes, we ran several models with the use of monthly averages of CDS spreads. There are some discrepancies in results, but the main findings do not change.

³An important point raised about the dynamics of a cointegrated system is that the conventional wisdom was incorrect (Enders 2004). If the linear relationship between two variables is already stationary, meaning that they are cointegrated, differencing the relationship entails a misspecification error.

⁴Granger causality test results (for the markets which are not cointegrated) are presented in Table 3.

⁵In general, α₂ should be positive and α₁ should be negative; therefore, α₂ – α₁ should be positive to correct the error from the previous period.